18.07.2023

Artificial intelligence (AI) and its use is on everyone's lips right now. How AI will change and shape our future is being hotly debated. AI applications are also trending in healthcare. But before they can deliver on their huge expectations, the basics have to be met. We spoke to three experts about the status quo of digitization and what role AI can or will play in hospitals in the future.

"The first step is always digitization. If you still write everything on a piece of paper, you don't need AI either"

Arne Peine, MD, Founder and CTO, Clinomic Group GmbH

There is already an incredible amount of data accumulating in a hospital today. Especially in the intensive care unit, multiple streams of data need to be processed. Since patients are monitored extensively here, medical staff have many values at their disposal in order to make a treatment decision that is both well-founded and individualized. Vital signs, laboratory values and the patient's medical history, including any interventions that may have been performed, must all be taken into account. The aspect of time also plays a role in the success of treatment. It is therefore important that physicians and nursing staff always have an overview of all the data – accessible in real time.

This is where Clinomic comes in. The start-up has tasked itself with "reinventing the hospital". With Mona, it has developed an intelligent assistance system that bundles all patient data and makes it available directly at the hospital bedside. Together with AI algorithms, Mona provides support for optimal treatment.

Ultimately, all the issues surrounding digitization and data use, as well as the use of AI, are not just about relieving the burden on hospital staff. The focus is also on the well-being of patients. After all, better digitization also has an impact on the quality of treatment.

Arne Peine, founder and CTO of Clinomic states that digitization alone is not enough and insufficient to just to generate data. It's also about sensible data management. In the intensive care unit, about 1,000 data points are generated per patient per hour. This quantity needs to be processed and evaluated. But data security is also a point that should not be neglected.

Products and exhibitors on this topic

Discover relevant products and exhibitors in the MEDICA database:

Further exciting articles from the MEDICA-tradefair.com editorial team:

Data Mining

In the video, Dr. Peine talks about the health data that accumulates and how it can be used sustainably with data mining.

Tobias Strapatsas, MD

"AI is only one aspect"

What can already work for the intensive care unit due to the "automatically" accruing data, Tobias Strapatsas, MD, would also like to see for his area of activity. Strapatsas initiated the project KIBATIN as head physician of the central emergency room at the Mönchengladbach Municipal Hospital. Here, too, time is of the essence. Medical and nursing staff must assess within minutes how urgently the arriving person needs medical care, even before the question of how exactly to treat him or her can be resolved. Determining treatment priorities in the course of triaging takes place in the emergency department as a standard procedure with the help of tried-and-tested initial assessment tools. For patients who come to the emergency department on their own, the data available for diagnosis and triage is usually very sparse.

The situation is different for "patients who come with the ambulance service. Experience shows that we already have much more information. Namely, everything that the ambulance service has already found out or initiated in terms of treatment measures. However, hospitals in general do not make sufficient use of this information in the context of triage," stresses Dr. Strapatsas, who has been head physician of the central emergency department at the Asklepios Hospital Hamburg Harburg since June 2023.

And this is precisely where the funded project "KI-basiertes Assistenzsystem für eine zuverlässigere Priorisierung in der Notaufnahme" (KIBATIN; English: AI-based assistance system for more reliable prioritization in the emergency room) comes in.

"In the end, we want a system that makes our nursing work easier by pointing out critical points from the rescue service and also is already deriving suggestions based on this."

The problem is that not every piece of electronic information can be used in a structured way to feed an algorithm. Then there is data that is hidden in free text, but that also contains important information. In the end, it's like a jigsaw puzzle: only the totality of the pieces makes a picture.

"AI is only one aspect of the whole," Dr. Strapatsas said. "It doesn't take artificial learning to derive urgency from a check on the patient’s ‘ventilation’ item. Such an emergency, when the ambulance arrives, needs a qualified team standing by, which takes over further treatment." It is more a matter of training the algorithm for less clearly structured and rule-based cases. For this, the developers are using synthetic data.

KIBATIN

The researchers of the project are developing an assistance system to support and optimize the initial medical assessment in the emergency room. It analyzes data collected by the ambulance service. In doing so, it identifies relevant information and suggests a prioritization of emergencies with a comprehensible justification. A hybrid AI assistance system will be developed, based on human expertise and a data-driven component. The rule-based control provides high precision and reliability. The data-driven approach, on the other hand, increases coverage and assurance of results and is trained using machine learning techniques.

Synthetic data

Data points randomly generated by an algorithm and thus artificially created, but structured and built in the same way as the real data set. This means that it is no longer possible to draw conclusions about patients' private health data.

Nosocomial infection

Nosocomial infections are infections that are temporally related to a hospital stay, i.e., they usually occur only 48 hours after hospitalization. For this reason, they are also colloquially referred to as "hospital-acquired infections." In most cases, these are urinary tract, wound or respiratory tract infections. In the worst case, there is a risk of sepsis. The cause of these infections can be a lack of hygiene measures. But risk factors such as advanced age or chronic illnesses can also promote infections caused by pathogens.

A very similarly functioning system is used at the Sisters of Mercy Hospital in Ried, Austria. Here, however, the AI detects hospital infections.

HAIDI

HAIDI is an AI-based system from the Czech start-up Datlowe, which scans all available digital data from the hospital information system and compares it with internationally recognized criteria such as the KISS (Hospital Infection Surveillance System) criteria for nosocomial infections. From this, the system generates a list of patients potentially at risk, enabling near real-time tracking.

Dr. Milo Halabi, a member of the hospital's hygiene team, is excited: "Previously, we had to more or less chase down nosocomial infections or they were reported to us. Now we're able to prospectively and even interactively see who's developing an infection." That's also possible because HAIDI can detect clues in free text as well, such as notes from physicians and nurses.

The system was tested in Ried before it was officially introduced. And it has proven its worth. The employees take a look at the software and are shown all potential cases of infection. They then check these. This takes between 30 and 45 minutes. But the system has proved extremely accurate so far. For Dr. Halabi and his team, the Czech start-up's product was therefore a stroke of luck.

Milo Halabi, MD

"There is so much data lying around in healthcare facilities that could add immense value to treatment when linked and properly processed."

Milo Halabi, MD, specialist in clinical pathology and molecular pathology, site manager of the Institute of Pathology, Hospital of the Sisters of Mercy, Ried, Austria.

For this, it is important to adequately implement digitization and data management. Because even "HAIDI can only be as good as the digital architecture in a hospital is." The higher the level of digitization, the more usable data there is, the better results intelligent systems can deliver, he says. And in the end, everyone benefits. "Studies show for Austria that treating nosocomial infections costs up to 20 million euros a year." So detecting hospital-acquired infections more quickly also saves resources for diagnostics, medications, personnel and in terms of bed capacity.

But what will the hospital of the future look like?

All three interviewees agree that, despite all technical progress, the well-being of patients is paramount.

Digitization always brings challenges, and not only in the handling of data, but also in the development and implementation of meaningful processes. But the use of AI is proving to be indispensable in a growing array of applications.

Clinomic has therefore made it its mission to actively shape the digital transformation of healthcare. For founder and CTO, Arne Peine, humans and computers harmonize well when both play to their strengths - treatment and care on the one hand, and analytical data evaluation for intelligent forecasts and predictions on the other.

For Dr. Strapatsas, however, the hospital of the future will not be as obviously equipped with AI systems as one might commonly think.

"My concern is that digitization is not adequately thought through to the end."

Tobias Strapatsas, MD, specialist in internal medicine, clinical acute and emergency medicine, Asklepios Hospital Hambur Harburg (until June 2023 Mönchengladbach Municipal Hospital)

"If AI is understood as a big system that makes decisions or gives opinions on X-rays, then I think we will have such applications. But I don't think those will be the vast majority. The vast majority of AI applications will be at many points where we won't even notice them."

At the hospital in Ried, Halabi is also positive about the opportunities that digitization and AI will bring. "My team and I have a great affinity for digital solutions." So HAIDI will certainly not be the last assistance system to be used there.